Advancing Local Clustering on Graphs via Compressive Sensing: Semi-supervised and Unsupervised Methods
Researchers have developed new methods for local clustering on graphs, focusing on identifying substructures within large, unlabeled datasets. The proposed techniques include a semi-supervised approach for scenarios with minimal labeled data and an extension to fully unsupervised settings. These methods involve graph sampling, diffusion processes, and overlap analysis to extract and verify local clusters, demonstrating state-of-the-art performance in low-label regimes. AI
IMPACT Introduces novel techniques for analyzing graph structures, potentially improving performance in areas like network analysis and recommendation systems.